A quality control system flags batches as defective if more than 3 out of 50 items are faulty. Each item has an independent 4% defect probability. Without computing the exact binomial CDF, which approach best approximates the probability of a batch being flagged?